Drive assistance device, drive assistance method and program

Abstract

To realize drive assistance related to a process in which continuous information and discrete information exist.SOLUTION: A drive assistance device according to one embodiment of the present disclosure has: an input part that inputs first process data including continuous information and discrete information; a storage part that stores a plurality of pieces of second process data including the continuous information and the discrete information; a classification part that classifies the first process data and the plurality of pieces of second data into a plurality of groups based on the discrete information; a similarity search part that searches for the second process data similar to the first process data from among one ore more pieces of second process data classified into the same group as that of the first process data based on the continuous information; and a correction part that corrects an amount of operation included in the second process data similar to the first process data.SELECTED DRAWING: Figure 4

Description

[Technical field]

[0001] The present disclosure relates to a driving assistance device, a driving assistance method, and a program. [Background technology]

[0002] In recent years, the need for plant operation support technology has been increasing in the industrial field due to a shortage of skilled operators, and various technologies related to plant operation support have been proposed. For example, Patent Document 1 discloses a technology that targets a plant that executes a continuous process, searches for process data of past operations similar to the current operating state, determines operation variables from the process data, and presents them to an operator, etc. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 5176206 Summary of the Invention [Problem to be solved by the invention]

[0004] However, in conventional technology related to plant operation support, process data of past operations is searched for from information that represents continuous values ​​of a process (hereinafter also referred to as continuous information). Therefore, when information that represents discrete values ​​of a process (hereinafter also referred to as discrete information) exists in addition to the continuous information, it is sometimes impossible to search for process data of past operations in a state similar to the current operating state.

[0005] The present disclosure has been made in consideration of the above points, and aims to realize operational assistance for a process in which continuous information and discrete information exist. [Means for solving the problem]

[0006] A driving assistance device according to one aspect of the present disclosure includes an input unit that inputs first process data including continuous information and discrete information, a storage unit that stores a plurality of second process data including the continuous information and the discrete information, a classification unit that classifies the first process data and the plurality of second process data into a plurality of groups based on the discrete information, a similarity search unit that searches for second process data similar to the first process data from one or more second process data classified into the same group as the first process data based on the continuous information, and a correction unit that corrects an operation amount included in the second process data similar to the first process data. Effect of the Invention

[0007] It is possible to realize operational assistance for processes in which both continuous and discrete information exist. [Brief description of the drawings]

[0008] [Figure 1] FIG. 2 is a diagram illustrating an example of a hardware configuration of the driving assistance device according to the present embodiment. [Diagram 2] FIG. 2 is a diagram illustrating an example of a functional configuration of a driving assistance device according to the present embodiment. [Diagram 3] FIG. 13 is a diagram showing an example of batch process data of past operations. [Figure 4] 5 is a flowchart illustrating an example of a driving assistance process according to the present embodiment. [Diagram 5] FIG. 13 is a diagram showing an example of classification of target data and batch process data of past operations. [Figure 6] FIG. 13 is a diagram illustrating an example of a neighborhood search. [Figure 7] FIG. 13 is a diagram showing an example of correction to an operation amount of a similar operation and calculation of an evaluation item value. [Figure 8] FIG. 13 is a diagram illustrating an example of a visualization result of recommended operation amounts. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] An embodiment of the present invention will be described below. In the following embodiment, a batch process is mainly assumed as a process in which continuous information and discrete information exist, and an operation support device 10 that realizes operation support for a plant that executes a batch process will be described. Note that continuous information is information that represents continuous values ​​of a process. On the other hand, discrete information is information that represents discrete values ​​of a process, and takes a constant value without changing within the same operation.

[0010] Examples of continuous information include measured values ​​(specific examples: temperature, pressure, flow rate, etc.) obtained by measuring the state of a process using a sensor or the like. Examples of discrete information include information indicating the name of a product manufactured or produced by a process, and information indicating the names of raw materials or materials input to a plant that executes the process. In general, process data of a batch process (hereinafter also referred to as batch process data) is composed of continuous information such as measured values ​​and some discrete information (for example, information indicating the name of a product manufactured or produced by the batch process or the names of raw materials input to a plant that executes the batch process). However, the discrete information does not change within the batch process (i.e., the same operation) and takes a constant value.

[0011] Note that the discrete information is not limited to the above as long as it is information expressed as discrete values, and may be, for example, information indicating the number of a furnace in which a process is executed, information indicating the number of a tank in which raw materials and the like to be input to a plant are stored, etc. Furthermore, the discrete information may be, for example, information indicating the month, day of the week, time period, season, etc. in which a process is executed, information indicating the execution time of a process, information indicating the weather (sunny, rainy, cloudy, etc.) on the day a process is executed, etc. As another example, the discrete information may be information indicating values ​​obtained by discretizing some continuous information using a technique such as clustering.

[0012] Here, the operation support device 10 according to this embodiment realizes operation support for a plant that executes a batch process by executing the following (1) to (4).

[0013] (1) Using discrete information, process data of current operations and process data of past operations are classified into multiple groups.

[0014] (2) Using the continuity information, process data of past operations (hereinafter also referred to as similar operations) in a state similar to the current operation state is obtained from the process data classified into the same group as the process data of the current operation.

[0015] (3) One or more manipulated variables are generated by performing corrections on the manipulated variables included in the process data acquired in (2) above.

[0016] (4) Among the manipulated variables corrected in (3) above, the manipulated variables that make a specified evaluation item value (e.g., an item value for evaluating the quality of a product produced or manufactured by a batch process) reach a target value are visualized and presented to an operator, etc. as recommended manipulated variables.

[0017] The above (1) to (4) realize operation assistance for a process in which continuous information and discrete information exist. In particular, the above (1) and (2) make it possible to acquire process data of a similar operation by considering both continuous information and discrete information. Furthermore, the above (3) makes it possible to obtain an optimal manipulated variable that achieves a target even if the manipulated variable of the current operation and the manipulated variable of the similar operation are not very similar. Furthermore, the above (4) makes it possible for an operator to determine and set an appropriate manipulated variable by referring to the recommended manipulated variable when actually setting the manipulated variable in the plant (more precisely, when setting the manipulated variable in a control device that controls a plant including the equipment to be operated).

[0018] As an example, the batch process data is a set of N process variables x 1 ,x N , and the operational variable u, which can take continuous or discrete values, and the K attribute values ​​a, which can take discrete values. 1 ,···,a KFor simplicity, the following will be described assuming that K=2. In this case, the batch process data is composed of {(x 1 (t),...,x N (t), u(t), a 1 (t),a 2 (t))|t=1, ,T}, where the values ​​of N process variables (x 1 (t),...,x N (t) is continuous information, and two attribute values ​​a 1 (t),a 2 (t) is the discrete information. The value of the manipulated variable, u(t), is the manipulated variable. Note that t is an index that represents a relative time within a batch.

[0019] In the following, the process variable x at each time t is 1 ,x N The value of the operation variable u and the attribute value a 1 ,a 2 The set of (x) and (x) is called the "batch process data at time t". 1 (t),...,x N (t), u(t), a 1 (t),a 2 Let us call the data (t) the batch process data at time t.

[0020] <Hardware configuration example of the driving support device 10> An example of a hardware configuration of a driving support device 10 according to this embodiment is shown in Fig. 1. As shown in Fig. 1, the driving support device 10 according to this embodiment has an input device 101, a display device 102, an external I / F 103, a communication I / F 104, a RAM (Random Access Memory) 105, a ROM (Read Only Memory) 106, an auxiliary storage device 107, and a processor 108. Each of these pieces of hardware is connected to each other via a bus 109 so as to be able to communicate with each other.

[0021] The input device 101 is, for example, a keyboard, a mouse, a touch panel, a physical button, etc. The display device 102 is, for example, a display, a display panel, etc. Note that the driving assistance device 10 does not necessarily have to include at least one of the input device 101 and the display device 102, for example.

[0022] The external I / F 103 is an interface with an external device such as a recording medium 103a. Examples of the recording medium 103a include a CD (Compact Disc), a DVD (Digital Versatile Disk), an SD memory card (Secure Digital memory card), and a USB (Universal Serial Bus) memory card.

[0023] The communication I / F 104 is an interface for connecting the driving assistance device 10 to a communication network. The RAM 105 is a volatile semiconductor memory (storage device) that temporarily stores programs and data. The ROM 106 is a non-volatile semiconductor memory (storage device) that can store programs and data even when the power is turned off. The auxiliary storage device 107 is a non-volatile storage device such as a hard disk drive (HDD), a solid state drive (SSD), or a flash memory, and stores programs and data. The processor 108 is various types of arithmetic devices such as a central processing unit (CPU) or a graphics processing unit (GPU).

[0024] 1 is an example, and the driving support device 10 may have other hardware configurations. For example, the driving support device 10 may have multiple auxiliary storage devices 107 and multiple processors 108, or may have various hardware other than the hardware shown in the figure.

[0025] <Example of functional configuration of driving assistance device 10> FIG. 2 shows an example of a functional configuration of the driving support device 10 according to the present embodiment. As shown in FIG. 2, the driving support device 10 according to the present embodiment includes an input unit 201, a classification unit 202, a similarity search unit 203, a correction unit 204, a selection unit 205, and a visualization unit 206. These units are realized, for example, by a process in which one or more programs installed in the driving support device 10 are executed by the processor 108 or the like. The driving support device 10 according to the present embodiment also includes a storage unit 207. The storage unit 207 is realized, for example, by a storage area of ​​the auxiliary storage device 107 or the like. The storage unit 207 may be realized, for example, by a storage area of ​​a storage device connected to the driving support device 10 so as to be able to communicate with the driving support device 10.

[0026] The input unit 201 inputs batch process data (hereinafter also referred to as target data) for calculating the recommended operation amount in the current operation (i.e., the batch process currently being executed). Here, the target data is represented by t=1, where t is the start time of the batch in the current operation, t=t is the current time, and t is the time of the current operation. p As {(x 1 (t),...,x N (t), u(t), a 1 (t),a 2 (t))|t=1, ,t p} In other words, the target data is the data from the start time t=1 of the batch in the current operation to the current time t=t p It is a set of batch process data from time t up to time t. Note that the input unit 201 may acquire the target data from, for example, a control device that controls a plant that is currently executing a batch process, or may acquire the target data from a database server that stores batch process data of the current operation.

[0027] The classification unit 202 uses discrete information contained in the target data and discrete information contained in the batch process data stored in the memory unit 207 to classify the target data and the batch process data stored in the memory unit 207 into multiple groups.

[0028] The similarity search unit 203 uses continuity information contained in the target data and continuity information contained in batch process data classified into the same group as the target data to search for and acquire batch process data of similar operations from among the batch process data.

[0029] The correction unit 204 creates one or more recommended manipulated variable candidates by correcting the manipulated variables included in the batch process data of similar operations. Here, the recommended manipulated variable candidates refer to a manipulated variable series (time-series data of the manipulated variables) that are candidates for the recommended manipulated variables.

[0030] The selection unit 205 selects the current time t=t p Subsequent operation amount (or time t>t p The evaluation item value for each recommended manipulation amount candidate is calculated using a prediction model that predicts a predetermined evaluation item value when a manipulation amount represented by the recommended manipulation amount candidate is set in the plant as a manipulation amount (hereinafter referred to as a "manipulation amount"). Then, the selection unit 205 selects the recommended manipulation amount candidate whose evaluation item value reaches a target value as the recommended manipulation amount. Here, the evaluation item is an item for evaluating a batch process, and examples thereof include items for evaluating the quality of a product produced or manufactured by the batch process. Also, the target value is a target value for the evaluation item, and examples thereof include a target value for product quality. Specific examples of product quality include, for example, the hardness of a tablet or the concentration of a certain component when a certain tablet is produced by a batch process.

[0031] However, the evaluation items being items for evaluating the quality of the product are just an example, and other items may be used as the evaluation items. For example, the number of products produced or manufactured by the batch process may be used as the evaluation item, or the time required to produce or manufacture the products may be used as the evaluation item. In addition, any item that evaluates the batch process may be used as the evaluation item.

[0032] Any model can be used as the prediction model as long as it predicts an evaluation item value when a manipulation quantity represented by a recommended manipulation quantity candidate is set in a plant based on the recommended manipulation quantity candidate. For example, a machine learning model that inputs a series of manipulation quantities represented by the recommended manipulation quantities (or some feature quantity such as a statistic of the series) and outputs an evaluation item value can be used as the prediction model.

[0033] The visualization unit 206 visualizes at least the recommended operation amount selected by the selection unit 205. At this time, the visualization unit 206 may visualize the recommended operation amount on the display device 102, for example, or on a display provided in a terminal (particularly, a terminal used by an operator) communicably connected to the driving support device 10.

[0034] The storage unit 207 stores batch process data of past operations. When each operation is completed (i.e., when each batch is completed), the batch process data of the operation is stored in the storage unit 207 as batch process data of the past operation.

[0035] Here, an example of the batch process data of the past operation stored in the storage unit 207 is shown in FIG. 3. In the example shown in FIG. 3, the batch process data of the past operation 1 {(x 1 (t),...,x N (t), u(t), a 1 (t),a 2 (t)|t=1, ,T 1}~Batch process data of past operation M{(x 1 (t),...,x N (t), u(t), a 1 (t),a 2 (t)|t=1, ,T M3, M batch process data of past operations m (m∈{1, ,M}) are stored in the storage unit 207. Note that, although the batch process data of each past operation represents one batch of process data, the length of the batch may differ for each past operation, and therefore the end time of the batch may also differ for each past operation. For this reason, in the example shown in FIG. 3, the end time of the batch of past operation m (m∈{1, ,M}) is t=T m In addition, the batch process data for each past operation m {(x 1 (t),...,x N (t), u(t), a 1 (t),a 2 (t)|t=1, ,T m}, discrete information a 1 (t),a 2 Note that (t) is a constant.

[0036] <Driving assistance processing> The operation support process according to this embodiment will be described below with reference to Fig. 4. Note that the operation support process described below may be executed repeatedly every time a predetermined time elapses, may be executed at a predetermined timing, or may be executed when a predetermined event occurs in a batch process. Examples of the event include an event indicating the start of a batch, an event indicating that an operation amount of some operation has been set in the plant, etc.

[0037] Step S101: The input unit 201 inputs target data (i.e., the start time t=1 of the batch in the current operation to the current time t=t p A set of batch process data from time t to time t is input.

[0038] Step S102: The classification unit 202 classifies the target data and each batch process data into a plurality of groups using the discrete information included in the target data input in the above step S101 and the discrete information included in the batch process data stored in the storage unit 207. For example, the classification unit 202 may classify the target data and each batch process data of the past operations into a plurality of groups by treating data having the same discrete information as belonging to the same group.

[0039] As an example, if the discrete information is attribute value a 1 , a 2 and a 1 and a 2 Figure 5 shows an example of classification of target data and batch process data of past operations when P 0 is the attribute value of the target data (a 1 ,a 2 ) plotted, P 1 ~P 18 is the attribute value of the batch process data of past operations (a 1 ,a 2 ) are plotted. In this case, as shown in Figure 5, 1 ,a 2 The attribute values ​​of the target data are classified into four groups: (a 1 ,a 2 )=(0,0), the batch process data in the same group as the target data is at point P 1 ~P 6 These are the batch process data corresponding to each of the above.

[0040] Step S103: The similarity search unit 203 searches and acquires batch process data of similar operations from among the batch process data, using the sequential information included in the target data input in the above step S101 and the sequential information included in the batch process data classified into the same group as the target data. Here, the similarity search unit 203 may, for example, search and acquire batch process data of similar operations by using a time cross-section of the sequential information, or may search and acquire batch process data of similar operations by using a time series of sequential information for a certain period of time. These two methods will be described below as examples. However, these two methods are merely examples, and batch process data of similar operations may be searched and acquired by methods other than these two methods.

[0041] -Methods that utilize time slices of continuous information The target data is d, and the batch process data at time t included in the target data is d(t). Also, the batch process data of past operations classified in the same group as the target data d is d i , batch process data d i The batch process data at time t contained in i (t), where i represents the number of batch process data of past operations that is classified into the same group as the target data d, and i = 1, . . . , I. I is the total number of batch process data of past operations that is classified into the same group as the target data d.

[0042] In the following, for simplicity, it is assumed that only one type of operation is performed in one batch, and a case is considered in which it is desired to present the recommended operation amount of this operation to an operator, etc. If multiple types of operations are performed in one batch, the following can be similarly applied to the operation for which the recommended operation amount is desired to be presented to an operator, etc.

[0043] The feature quantity of the target data d is w:=d(t p -Δ), batch process data d i The feature of w i :=(d i (t i(1) -Δ)+···+d i (t i (Si) -Δ)) / Si, where t p is the current time, and Δ is a predetermined time span (for example, Δ = 1 minute). i (s) is the batch process data d i The time when the sth (s=1, , Si)th operation is performed in the batch process data d. In other words, the time when u(t-1) ≠ u(t) occurs due to the sth operation. i This is the number of times the operation was performed in the past operation corresponding to

[0044] In particular, Si=1 may be used. In this case, the batch process data d i The feature value of w i =d i (t i -Δ).

[0045] At this time, a method such as neighborhood search is used to find the feature value w i Among (i=1, ,I), the feature w that is most similar to feature w i Batch process data of i is obtained as batch process data for similar operations. Here, the feature quantity w i The scale for measuring the similarity between feature quantity w and feature quantity w is not limited to a specific scale, and any known scale may be used. For example, distance (specific examples: Euclidean distance, Manhattan distance, etc.) and similarity (specific examples: cosine similarity, Jaccard coefficient, etc.) may be used. In addition to distance and similarity, for example, correlation coefficient may be used.

[0046] As an example, N=2 and the feature quantity w i An example of a neighborhood search when Euclidean distance is used as a measure of similarity between and feature value w is shown in Figure 6. In the example shown in Figure 6, 0 is the feature value of the target data d w=d(t p -Δ) plotted point, P 1 ~P 6is the batch process data of past operations. i The feature value w i =(d i (t i (1) -Δ)+···+d i (t i (Si) -Δ)) / Si. In this case, point P 0 The nearest point of 3 Therefore, the batch process data d 3 is obtained as batch process data for similar operations.

[0047] -Methods using time series of continuous information over a certain period of time The target data is d, and the batch process data at time t included in the target data is d(t). Also, the batch process data of past operations classified in the same group as the target data d is d i , batch process data d i The batch process data at time t contained in i (t), where i represents the number of batch process data of past operations that is classified into the same group as the target data d, and i = 1, . . . , I. I is the total number of batch process data of past operations that is classified into the same group as the target data d.

[0048] For simplicity, the following assumes that one type of operation is performed only once in one batch, and considers a case where it is desired to present the recommended operation amount for this operation to an operator, etc. If multiple types of operations are each performed only once in one batch, the following can be similarly applied to the operations for which it is desired to present the recommended operation amount to an operator, etc.

[0049] The feature of the target data d is w, and the feature w is expressed as the time series data d(1), , d(t p -Δ) as the feature quantity. Similarly, the batch process data d i The feature of w i Then, the feature value w i Let d be the time series data i (1),...,d i(t i -Δ) feature. Here, t p is the current time, and Δ is a predetermined time span (for example, Δ = 1 minute). i is the batch process data d i is the time when the corresponding operation is performed (in other words, the time t when u(t-1) ≠ u(t) as a result of the corresponding operation being performed).

[0050] Various feature quantities can be used as the feature quantity of time series data, for example, statistics of time series data (specific examples: average value, variance, etc.), the sum of distances between adjacent data included in the time series data (i.e., the distance between data at time t-1 and data at time t), etc. Regarding the distance between adjacent data, for example, Euclidean distance, Manhattan distance, Dynamic Time Wrapping, etc. can be used.

[0051] At this time, a method such as neighborhood search is used to find the feature value w i Among (i=1, ,I), the feature w that is most similar to feature w i Batch process data of i is obtained as batch process data for similar operations. Here, the feature quantity w i The scale for measuring the similarity between feature quantity w and feature quantity w is not limited to a specific scale, and any known scale may be used. For example, distance (specific examples: Euclidean distance, Manhattan distance, etc.) and similarity (specific examples: cosine similarity, Jaccard coefficient, etc.) may be used. In addition to distance and similarity, for example, correlation coefficient may be used.

[0052] Step S104: The correction unit 204 creates one or more recommended operation amount candidates by correcting the operation amount included in the batch process data acquired in the above step S103. Here, the correction unit 204 may correct the operation amount by using each of a plurality of predefined correction patterns, may correct the operation amount by feedback control with respect to a given target value, may correct the operation amount so as to reach a given target value by model predictive control using a predictive model, may correct the operation amount so as to maximize a gain by using reinforcement learning, or may correct the operation amount by a method that combines these correction methods. Hereinafter, as an example, a method of correcting the operation amount by using each of a plurality of correction patterns will be described. However, all of the above correction methods are merely examples, and the operation amount may be corrected by various correction methods other than the above correction methods.

[0053] Hereinafter, the batch process data acquired in step S103 above will be referred to as d i' ={(x 1 (i') (t),...,x N (i') (t),u (i') (t),a 1 (i') (t),a 2 (i') (t)|t=1, ,T (i')}. In addition, the number of correction patterns is assumed to be J, and each correction pattern is represented as "correction pattern j". At this time, the correction unit 204 uses the correction pattern j to calculate the batch process data d i' The control variable u included in (i') (t) (where t p +1≦t≦T (i') ), the jth recommended operation variable candidate {u j (t)|t=1, ,T (i')} where 1≦t≦t p When j (t)=u (i') (t), t p +1 ≤ t ≤ T', then u j (t)=hj (u (i') (t)), where h j (·) is a function that takes the correction amount as input, corrects the operation amount by the correction pattern j, and outputs the corrected operation amount. That is, h j (·) is a function that represents the correction pattern j. This allows the J number of recommended operation amount candidates {{u j (t)|t=1, ,T (i')}|j=1,···,J} is obtained.

[0054] The above function h j (·) is created in advance and stored in, for example, the auxiliary storage device 107. j A specific example of (·) is t 1 ≦t <t 2 When h j (u(t))=u(t), t 2 ≦t <t 3 When h j (u(t))=a×u(t), t 3 ≦t <t 4 When h j (u(t)) = b × u(t), t 4 When t≦h j Functions defined as (u(t))=c, etc., where t 1 ~t 4 is an index that represents a preset time (relative time within a batch), and a, b, and c are preset constants. j (u(t))=a×u(t)+b (where a and b are preset constants). However, these are just examples. j (·) can be defined in various ways. For example, any polynomial function or elementary function (trigonometric function, exponential function, logarithmic function, etc.) can be combined into h j (·), or a function that assigns random numbers to the manipulated variable can be further combined to define h j It can also be defined as (·). It is assumed that J≧2, but J=1 is also acceptable.

[0055] Step S105: The selection unit 205 uses the prediction model to select recommended operation amount candidates to be used as the recommended operation amounts from the recommended operation amount candidates created in the above step S104.

[0056] The recommended operation amount candidates created in step S104 above are {{u j (t)|t=1, ,T (i')}|j=1, . . . , J}. In addition, the prediction model is f. In this case, the selection unit 205 selects, for example, {u j (t)|t=t p ,···,T (i')}, the evaluation item value E j Then, the selection unit 205 calculates the (i') Evaluation item value E in j (T (i') ) reaches the given target value V as the recommended operation amount. That is, the selection unit 205 selects, for example, the evaluation item value E j (T (i') ) and the target value V is |E j (T (i') )-V|<ε and |E j (T (i') The recommended operation amount candidate with the smallest |E |-V| is selected as the recommended operation amount. Here, ε>0 is the allowable error. In this way, the optimal operation amount that allows the evaluation item value to reach the target value within the allowable error can be obtained as the recommended operation amount. However, this is only an example, and the selection unit 205 may select, for example, |E j (T (i') )-V| may be selected as the recommended manipulated variable.

[0057] The prediction model f is based on the time series data of the operation variables included in the recommended operation variable candidates {u j (t)|t=t p ,···,T (i')}, or a part of the time series data, or a feature of the time series data. (i') Evaluation item value E inj (T (i') ) or the time series data of the evaluation items {E j (t)|t=t p ,···,T (i')}.

[0058] As an example, FIG. 7 shows a case where the operation amount is corrected using the correction pattern 1 to the correction pattern K in the above step S104, and the evaluation item value is calculated by the prediction model in the above step S105. In the example shown in FIG. 7, the evaluation item value predicted from the recommended operation amount candidates created using the correction pattern 1 is (i') The target value is reached at . Therefore, in the example shown in Fig. 7, the recommended manipulated variable candidates created by correction pattern 1 are selected as the recommended manipulated variable. Note that, in the example shown in Fig. 7, the evaluation item values ​​are shown as achievement rates [%] relative to the target value.

[0059] Step S106: The visualization unit 206 visualizes at least the recommended operation variables selected in the above step S105. Note that, in addition to the recommended operation variables, the visualization unit 206 may visualize, for example, time series data of evaluation item values ​​corresponding to the recommended operation variables, time series data of operation variables included in the batch process data of a similar operation, time series data of evaluation item values ​​calculated from operation variables included in the batch process data of a similar operation, etc.

[0060] As an example, FIG. 8 shows a result of visualizing the recommended manipulated variables, as well as the time series data of the evaluation item values ​​corresponding to the recommended manipulated variables, the time series data of the manipulated variables included in the batch process data of a similar operation, and the time series data of the evaluation item values ​​calculated from the manipulated variables included in the batch process data of a similar operation. The visualization result 1000 shown in FIG. 8 includes an operation plan field 1100 and a similar operation field 1200. The operation plan field 1100 includes a recommended manipulated variable display field 1110 in which a graph visualizing the recommended manipulated variables is displayed, and an evaluation item display field 1120 in which a graph visualizing the time series data of the evaluation item values ​​corresponding to the recommended manipulated variables is displayed. The similar operation field 1200 includes a similar operation manipulated variable display field 1210 in which a graph visualizing (the time series data of) the manipulated variables included in the batch process data of a similar operation is displayed, and an evaluation item display field 1220 in which a graph visualizing the time series data of the evaluation item values ​​calculated from the manipulated variables is displayed. This allows an operator or the like to know the operation amount that will cause the evaluation item value to reach the target value in the current operation as the recommended operation amount.

[0061] 8 may include a selection field 1300. The selection field 1300 includes radio buttons 1310 and a decision button 1320, and the radio buttons 1310 allow the selection of "recommended operation amount," "operation amount of similar operation," or "not adopted." The operator, etc., can select "recommended operation amount" from the radio buttons 1310 and then press the decision button 1320 to select the operation amount that is the most suitable for the current time t p Similarly, the operator can set the manipulated variable for the plant from the current time t among the manipulated variables included in the batch process data of the similar operation by selecting the “manipulated variable for similar operation” from the radio button 1310 and pressing the decision button 1320. pThe operator can set the manipulated variable from the recommended operation to the plant by selecting "Do not adopt" from the radio button 1310 and then pressing the decision button 1320. Setting the manipulated variable to the plant means, more precisely, setting the manipulated variable to a control device that controls the plant including the equipment to be operated.

[0062] <Summary> As described above, the operation support device 10 according to this embodiment acquires process data of past operations similar to the current operation while considering both continuous information and discrete information, and presents the correction amount obtained by correcting the operation amount included in this process data to the plant operator, etc. This makes it possible to prevent a situation in which process data of past operations that is not similar to the current operation is acquired. In addition, since the operation amount included in the process data of the past operation is corrected, even if the operation amount included in the process data of the past operation is not very similar to the operation amount of the current operation, the optimal operation amount that achieves the goal can be presented to the operator, etc.

[0063] <Modification> A modification of the above embodiment will now be described.

[0064] <Variation 1> 4, when searching for batch process data of past operations by neighborhood search, one batch process data that is closest to the target data is acquired as batch process data of similar operations, but multiple batch process data may be acquired. For example, in addition to the batch process data that is closest to the target data, the next closest batch process data and the next closest batch process data may be acquired.

[0065] In addition, when a plurality of batch process data are acquired as batch process data of similar operations, step S105 is executed for each batch process data of similar operations to select a plurality of recommended operation amounts, and then these plurality of recommended operation amounts are visualized in step S106. However, for example, when a plurality of batch process data acquired as similar operations are batch process data related to batch processes of equal length, one batch process data may be created by calculating an average value or the like of the plurality of batch process data at each time point, and then steps S105 to S106 may be executed for this batch process data.

[0066] <<Variation 2>> 4, the target data and each batch process data are classified into a plurality of groups, but batch process data having the same discrete information as the target data may be identified from the batch process data stored in the storage unit 207. In this case, in step S103, batch process data of similar operations may be searched for and acquired from the batch process data identified in step S102.

[0067] <Modification 3> In the above embodiment, a batch process is assumed as a process in which continuous information and discrete information exist, but the above embodiment may be applied to process data of a continuous system process divided into a certain period as long as the process has continuous information and discrete information. For example, the above embodiment may be applied to data obtained by dividing the process data of a continuous system process into periods of one month. Alternatively, the above embodiment may be applied to data obtained by dividing the process data of a continuous system process into periods of one month and adding discrete information (i.e., discrete information taking any value from 1 to 12) representing the month of the period.

[0068] The present invention is not limited to the above-described embodiments specifically disclosed, and various modifications, changes, combinations with known technologies, etc. are possible without departing from the scope of the claims. [Explanation of symbols]

[0069] 10 Driving assistance devices 101 Input Device 102 Display device 103 External I / F 103a Recording media 104 Communication I / F 105 RAM 106 ROM 107 Auxiliary storage 108 processors 109 Bus 201 Input section 202 Classification Department 203 Similarity Search Department 204 Correction section 205 Selection section 206 Visualization section 207 Memory section

Claims

1. an input unit for inputting first process data including continuous information and discrete information; a storage unit configured to store a plurality of second process data including the continuous information and the discrete information; a classification unit that classifies the first process data and the plurality of second process data into a plurality of groups based on the discrete information; a similarity search unit that searches for second process data similar to the first process data from one or more second process data classified into the same group as the first process data based on the continuity information; a correction unit that corrects an operation amount included in second process data similar to the first process data; A driving assistance device having the above configuration.

2. The discrete information is one or more attribute values ​​each taking a discrete value; The classification unit includes: The driving assistance device according to claim 1 , wherein the first process data and the plurality of second process data are classified into a plurality of groups by classifying data having the same one or more attribute values ​​into the same group.

3. The similarity search unit is 3. The driving assistance device according to claim 1, further comprising: a neighborhood search using a distance or a similarity between a feature amount of the sequential information included in the first process data and a feature amount of the sequential information included in each of the one or more second process data, to search for second process data similar to the first process data.

4. The driving assistance device according to claim 3 , wherein the feature amount of the continuous information is the feature amount or the statistic amount of the continuous information at a predetermined time cross section included in the time series of the continuous information, or the feature amount or the statistic amount of a predetermined time section in the time series of the continuous information.

5. The driving assistance device according to claim 4 , wherein the continuous information is one or more process variable values ​​included in the first process data and the second process data, each of which takes a continuous value.

6. The driving assistance device according to claim 5 , wherein the first process data and the second process data are process data of a batch process.

7. The correction unit is The driving support device according to claim 1 , further comprising: a control unit for controlling a driving force of the driving force control device, the control unit being configured to control a driving force of the driving force control device in accordance with a predetermined correction method;

8. The driving support device according to claim 7 , further comprising a selection unit that selects an optimal operation amount for achieving a predetermined target, based on the one or more corrected operation amounts.

9. The correction unit is creating one or more corrected manipulated variables by correcting the correction amounts after the current time among the manipulated variables included in second process data similar to the first process data based on one or more correction patterns; The selection unit is 9. The driving assistance device according to claim 8, further comprising: a control unit for controlling a driving force of the vehicle that is driven by the control unit and that is configured to control the vehicle to a desired target value, the control unit being ...

10. The selection unit is calculating a plurality of evaluation item values ​​from each of the one or more manipulated variable series using a model that receives the manipulated variable series as an input and outputs a predetermined evaluation item value; The driving support device according to claim 9 , wherein a series of operation amounts in which the evaluation item value reaches the target is selected as the optimal operation amount.

11. an input step of inputting first process data including continuous information and discrete information; a storage step of storing a plurality of second process data including the continuous information and the discrete information in a storage unit; a classification step of classifying the first process data and the plurality of second process data into a plurality of groups based on the discrete information; a similarity search procedure for searching for second process data similar to the first process data from one or more second process data classified into the same group as the first process data based on the continuity information; a correction step of correcting an operation amount included in second process data similar to the first process data; A driving assistance method implemented by a computer.

12. an input step of inputting first process data including continuous information and discrete information; a storage step of storing a plurality of second process data including the continuous information and the discrete information in a storage unit; a classification step of classifying the first process data and the plurality of second process data into a plurality of groups based on the discrete information; a similarity search procedure for searching for second process data similar to the first process data from one or more second process data classified into the same group as the first process data based on the continuity information; a correction step of correcting an operation amount included in second process data similar to the first process data; A program that causes a computer to execute the following.

Images